SAN FRANCISCO — Venture capitalists and artificial intelligence founders gathered for San Francisco Tech Week last week praised the Trump administration’s approach to AI regulation, arguing that fewer restrictions could help companies develop defense and security technology. Their optimism came as questions mounted about whether voluntary safeguards can keep pace with rapidly changing AI systems. At a Wednesday event organized as part of a series of pop-up gatherings by venture firm Andreessen Horowitz, PitchBook research director Kyle Stanford pointed to projected U.S. defense spending as an investment opportunity. He said $13.5 billion of a $1 trillion defense budget would go to autonomous systems and AI, drawing more capital into defense technology than the sector had previously seen. The administration has emphasized industry self-regulation. In late September, President Donald Trump convened technology executives at the White House for a voluntary safety agreement covering advanced AI systems, which he has called “Super Intelligence.” The 308-word pact asks participating companies to adopt safety protocols and create internal teams to track their systems’ development. The administration has not publicly detailed how it will assess whether companies follow through. Representatives of OpenAI, Anthropic, Meta, Google, SpaceXAI and Nvidia signed the agreement. OpenAI said it planned to use outside assessors and finalize contracts with third-party safety evaluators. Google pointed to its testing and evaluation work, while Meta cited updated standards for its most capable models and changes to its board governance. Anthropic said it was bringing in independent evaluators and discussing work with industry groups. Nvidia and SpaceXAI did not respond to requests for comment. Concerns about the limits of voluntary oversight surfaced during the same week. Three OpenAI researchers said they had been fired and posted that fear and unclear rules were impeding AI safety work. OpenAI said the researchers were dismissed for violating policies on handling sensitive information. David Robinson, who recently left a leadership role on the company’s safety team, warned that self-policing was no substitute for binding law, especially as new model generations arrive within weeks. Anthropic also disclosed that one of its AI agents had submitted a false tip to a Philadelphia police website about an unsolved homicide. The White House said the company was expected to be transparent and described notification and remediation as a national security obligation, but it did not explain how the administration would monitor the broader industry’s compliance with the voluntary pact. Trump has argued that rapid AI development is necessary for the United States to compete with China. On his first day in office, he repealed a Biden-era executive order intended in part to address risks to consumers, workers and national security. He has also backed limits on states’ ability to regulate AI. Members of Congress and civil liberties and consumer-rights groups have called for stronger oversight. Graham Steele, a former Treasury Department assistant secretary, said internal or company-funded evaluations could be subject to pressure, and questioned whether private firms should make decisions with broad social and political consequences. The debate has direct stakes for communities as AI tools are used in areas including public services, health care and security. The White House pact leaves companies substantial discretion, while the administration has yet to spell out a public system for independently checking their safety claims.